Go 1.25 introduces an experimental garbage collector named Green Tea that promises significant performance gains for many workloads. Already battle-tested at Google, this new collector can reduce garbage collection time by 10 to 40 percent. This article unpacks the ten most important things you need to know about Green Tea — from how to enable it to why it matters for your Go programs.
1. What Is the Green Tea Garbage Collector?
The Green Tea garbage collector is a new, experimental memory management component in Go 1.25. It replaces the existing concurrent mark‑sweep garbage collector with an algorithm that reduces CPU overhead during collection cycles. Unlike its predecessor, Green Tea is designed to minimize pause times and improve overall throughput, especially for applications that allocate memory aggressively. Early benchmarks show that many workloads experience around 10 percent less time spent in garbage collection, with some seeing reductions up to 40 percent. Because it is still experimental, developers must explicitly opt in by setting GOEXPERIMENT=greenteagc at build time.
2. Why the Name 'Green Tea'?
The name 'Green Tea' is a playful nod to both the color traditionally associated with Go's mascot and the idea of a 'lighter' garbage collection process — just as green tea is often perceived as a lighter beverage. In internal discussions, the team wanted a memorable, friendly moniker that conveys freshness and improved performance. The name also subtly hints at the collector's reduced resource footprint: it aims to 'brew' memory management more efficiently, leaving the CPU free to handle application logic. While the name is whimsical, the engineering behind it is solid and production‑proven.
3. How to Enable Green Tea GC in Go 1.25
To start using the Green Tea garbage collector, you need to build your Go toolchain or binaries with the GOEXPERIMENT=greenteagc environment variable. For example: GOEXPERIMENT=greenteagc go build -o myapp main.go. This flag activates the experimental runtime code that replaces the default collector. Note that Green Tea is not yet enabled by default; it will become the standard in Go 1.26 if all goes well. Because it's experimental, you should test your application thoroughly in a staging environment before deploying to production. The Go team provides clear instructions in the official documentation.
4. Performance Improvements You Can Expect
Performance gains vary by workload, but the Green Tea GC consistently reduces the amount of time the CPU spends on garbage collection. In typical web servers and API backends, developers report a 10 to 15 percent reduction in GC overhead. For memory‑intensive applications – such as large‑scale data processing or real‑time analytics – reductions can reach 40 percent. The improvement comes from better handling of large heaps and more efficient marking of live objects. However, not every workload benefits equally; some see negligible change, especially those that allocate very little or have extremely stable heap sizes.
5. Use at Google and Production Readiness
Google has been running the Green Tea garbage collector internally for months across a variety of services, from search infrastructure to cloud storage. According to the Go team, it is production‑ready and has handled billions of requests without incident. The collector passed rigorous stress tests and was tuned based on real‑world usage patterns. This internal validation gives the team confidence to encourage the community to try it out. Because it is already deployed at scale, you can trust that the core algorithms are stable and well‑tested.
6. The Roadmap: Default in Go 1.26
The Go team plans to make Green Tea the default garbage collector in Go 1.26, scheduled for release in late 2026. This timeline depends on community feedback and any unforeseen issues that may arise during the experimental phase. If your application runs smoothly with Green Tea in Go 1.25, it will seamlessly benefit from the new default in the next version. Conversely, if you encounter regressions, reporting them now gives the team time to address problems before the switch becomes permanent. The gradual rollout follows Go's tradition of careful, data‑driven evolution.
7. Where It Fits: Tracing Garbage Collection Basics
Understanding Green Tea requires a basic grasp of tracing garbage collection. Unlike reference counting, tracing collectors periodically pause the program to find objects that are no longer reachable. The collector starts from a set of root references (like global variables and stack frames) and follows pointers to mark all live objects. Unmarked objects are then reclaimed. Green Tea improves this process by reducing the time spent in the tracing phase and by allowing more concurrent work. It is still a tracing collector at its core, but with smarter algorithms that minimize disruption to application threads.
8. Understanding Objects and Pointers in Go
The garbage collector concerns itself with objects (heap‑allocated values) and pointers (references to those objects). In Go, heap objects are created when the compiler cannot allocate a value on the stack, for instance when escaping to a goroutine or when using new or make. Pointers link objects together, forming a directed graph. The collector traverses this graph to identify live objects. Green Tea's improvements come partly from better pointer tracking and reduced synchronization overhead during the mark phase. If you want to optimize your code for the new collector, focusing on pointer density and object lifetimes can help.
9. Workloads That Benefit Most (and Some That Don't)
Workloads with moderate to high allocation rates see the biggest gains with Green Tea – think microservices handling many concurrent requests or batch jobs that process large datasets. Conversely, workloads that are compute‑bound or have very small heaps may see little improvement. In some edge cases, applications that rely heavily on finalizers or that trigger frequent, large heap growth might even experience slight regressions. The Go team emphasizes that feedback is crucial: if Green Tea worsens your performance, please report it via the issue tracker. This helps refine the collector for all users.
10. How to Provide Feedback and Contribute
Your experience matters. If you try Green Tea and encounter bugs or performance issues, file a new issue on the Go issue tracker with details about your workload and Go version. If you see improvements, let the team know by replying to the existing Green Tea discussion thread. The community's reports help the Go team decide whether to make Green Tea the default and guide further optimizations. You can also contribute by running benchmarks with the collector enabled and sharing the results. The more data the team has, the better the final version will be for everyone.
The Green Tea garbage collector represents a major step forward for Go's runtime. It delivers tangible performance gains, is already proven at scale, and will become the default in Go 1.26. By trying it out in your own projects and giving feedback, you help shape the future of memory management in Go. So grab a cup of green tea (or your beverage of choice) and give GOEXPERIMENT=greenteagc a spin today!